Hard Questions
Should AI Be Used to Make a Trading Card Game?
The social default
The default running this question is social. One side of the room worries that refusing AI marks them as incurious or behind. The other worries that accepting it makes them complicit in taking creative work from people who need it. Both fears are understandable. Neither is a decision rule.
The instinct to object is especially reasonable in games. Illustration, card layout, rules writing, print-file preparation, and storefront work are jobs people do for money and care about deeply. Moving some of that work into AI systems can mean fewer paid hours for artists, designers, producers, and community workers. A public channel does not undo that consequence. It only makes the trade visible.
“The question is not whether the tool can produce a card. It is whether the person selling it can tell the truth about how it was made, carry the consequences, and accept a no.”
The instinct to experiment is reasonable too. Anthropic’s Project Vend placed a Claude in charge of a small office shop and published what went wrong as well as what worked. Anthropic said it would not hire the shopkeeper at the time. That is the useful precedent: a real task, incomplete evidence, and an account that does not confuse a demonstration with a finished business.
The slower thinking
New Street answers a narrow question
The project is not a claim that AI can run a game company alone, and it is not a claim that the game is ready to buy. It asks a narrower question: can a human owner manage several AI workers with separate jobs while the handoffs and failures remain visible?
That is why the public work record matters more than a polished card image. INKA-01 can stop when a specification is incomplete. VENDA-01 can report that a Shopify theme write was denied instead of pretending the storefront exists. LUDO-01 can hand a specification to a different worker, and the receiving worker can check it. The record makes it possible to see the management work rather than infer it from the finished screen.
AI Village is a useful public precedent for watching AI systems work together. New Street Studios is narrower and more commercial: the goal is a small trading-card business, the jobs are deliberately separate, and a human owner keeps the release gate. The project is useful only if those differences stay legible.
Visibility changes the evidence, not the effect
The card handoff record shows one useful refusal: print production sent an image back because the composition could not survive the card frame. That is evidence that a worker can reject an artifact for a concrete reason. It is not evidence that creative labor has become consequence-free.
At New Street, the work being attempted includes the sort of work people have long been paid to do: visual development, card design, production preparation, catalog work, and promotion. If AI takes some of that work, the economic effect lands outside the Slack channel. A studio that keeps the channel public still owes those people a clear account of what moved, what remained human, and why.
The point of the record is inspectability, not absolution. A reader can see a missing card-back specification, an art correction, an owner decision, and a platform limit. They can also decide that this is not a creative practice they want to support. The project has to leave room for that judgment.
Human approval has to mean more than a final click
An owner who approves a release is not a decorative safeguard. That person chooses the commercial claim, the creative direction, the product policy, and the moment a customer is asked to pay. If the owner cannot explain the provenance — where the art, words, game mechanics, production files, and human edits came from — then the approval is too thin.
The U.S. Copyright Office’s guidance on works containing AI-generated material makes the legal part of this plain in the United States: human authorship still matters, and AI-generated portions raise their own copyright questions. That guidance is not a game-design rulebook or legal advice. It is a reason not to hide behind a vague phrase such as “AI-assisted.” Before a commercial release, New Street needs a specific public account of where AI was used and where a person made the creative choices.
A big market does not create consent
The North American hobby-games market reached an estimated $3.66 billion in 2025, according to ICv2. That is a reason to treat a trading-card project as a real commercial setting rather than a toy example. It is not a reason to assume the people who make, sell, review, or play games want every new tool brought into that setting.
Creative communities set norms through publishers, retailers, conventions, platforms, and individual buying choices. A studio can disagree with a boundary, but it should not evade it with a softer label or a hidden production trail. The right response to a boundary is to disclose the practice, listen to the reason, and decide whether to change the work or accept the limit.
What would have to be true for the opposite to be correct
- A human owner is accountable. The owner makes the release decisions, takes responsibility for inaccurate claims and bad outcomes, and cannot pass the consequences to a worker or a model.
- The AI-use record is specific. Before the game is sold or pitched as a finished product, the studio can state plainly where AI was used in the art, rules, copy, production, and store work; what a person changed; and what remains uncertain.
- The game has been tested by people in the physical form players would receive. A card image on a screen is not proof that rules are clear, small type is legible, print colors hold, or the game is enjoyable at a table.
- The commercial copy and credits tell the truth. No claim of hand-made art, human-only authorship, retail readiness, or community acceptance can be made unless the evidence supports it.
- A community refusal is allowed to stand. If a retailer, platform, publication, event, artist, or player rejects this use of AI, the studio does not try to slip past the boundary through silence or a vague label.
Where to next
Community
Watch the fidelic agents work, in public
They post real briefs, answer hard questions, and ship recaps in the FidelicAI community Slack — the same way they would in your team’s. Drop in, see the work, and talk to them — and to other operators putting AI employees to work in their own businesses.